In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium – 100 Years ISPRS, Vienna, Austria, July 5–7, 2010, IAPRS, Vol. XXXVIII, Part 7B
Contents
Author Index
Keyword Index
RECOGNITION OF WINDING DISPLACEMENTS FOR STEEL COILS VIA LASER
LIGHT SECTION TECHNIQUE
Patrick A. Hölzl, Daniel C. H. Schleicher and Bernhard G. Zagar
Johannes Kepler University
Institute for Measurement Technology
Austria, Linz 4040
patrick.hoelzl@students.jku.at, daniel.schleicher@jku.at, bernhard.zagar@jku.at
KEY WORDS: Laser scanning, Measurement, Three–dimensional, Object, Pattern, Recognition, Segmentation, Industrial
ABSTRACT:
To satisfy the ever increasing quality standards of todays steel industry the basic products, in this case steel coils, must be produced
within very small tolerances. To achive those quality limits, control systems via machine vision are getting more and more popular. For
example the quality level of steel coils decreases due to winding displacements based on a non–ideal production process. In this paper
a machine vision system for the recognition of winding displacements of steel coils, based on the laser light sectioning technique, will
be presented. The introduction of the mathematical model of the laser light sectioning setup allows to compensate some influences of
the setups inaccuracies. We show that a reliable recognition of the coil profile with an accuracy below or less than 1mm can be achieved
by a rather simple adaptive algorithm. Finally defects can be detected accurately out of the recognized coil profile.
1
INTRODUCTION
Generally metal processing companies, like car manufacturer get
their basic material in form of rolled up steel sheets, called coils.
These metal processing companies and especially the steel industry are interested in systems to detect defects and to minimize material rejects. One of the major problems which can occur durring
the coil manufactoring process are displacements of windings,
due to a non–conform rolling–up process or a faulty packaging.
These winding displacements cause a higher percentage of non
usable material and so for quality aspects a preventive detection
of these displacements is necessary. Figure 1 shows a model of
the coil and winding displacements. The winding displacements
h(r) typically are less than 5 mm, larger displacements are problematic and are aimed to be detected by our system. To achieve
this requirement for the profil resolution the minimum detectable
winding displacement hmin (r) is specified to be equal to or less
than 1 mm.
Due to the first assumption we only need to look for defects on
one cylinder base. The second assumption limits the ROI to one
radial range of zero to dA /2 thus improving resolution. So for the
further recognition process the ROI will be limited to the lower
half of the coil as indicated in Fig. 2 (red rectangle). The last assumption is necessary for the recognition algorithm presented in
Sec. 4 to work correctly. The complete detection should happen
during the coils storage process, that means the coil will move
during the measurement. Figure 2 shows a typical storing process, with the coil on a transport waggon moving in axial direction with 0.5 m/s. The problem here is that the storage process
can not be influenced or changed, so the chosen measurement
method must be adapted to these conditions. Therfore an optical
3D measurement method is the best choice with respect to cost
and resolution. Due to the varying lighting conditions in the storage facility, which can not be influenced, the realized measurement system and furthermore the machine vision system must be
very robust with respect to natural illumination conditions.
Figure 1: Model of the coil with winding displacements h(r)
indicated, di and da are the inner and outer diameter of the coil
respectively and b is the mean coil width.
For the above–mentioned model we imply three assumptions which
are realistic for further processing:
1. An axial winding displacement h(r) on one side results in
an inverse displacement −h(r) on the other side (implying
b is constant for each winding).
2. The winding displacement h(r) is constant over one winding at radius r.
3. Winding displacements along the radius r occurs only in
small steps.
293
Figure 2: Moving steel coil on transport waggon during the storing process, with a blurred laser line barely visible in the ROI
(red rectangle).
Considering the given terms a detection system based on the laser
light sectioning technique (Kraus, 1996, Wu et al., 1993) is the
preferred choice. Figure 3 shows the principle behind the laser
light section technique. The elevation profile of interest is illuminated from an off-center position by a line shaped laser light fan.
So the lateral displacement d(r) of the projected line as viewed at
from a vantage point orthogonal to the front side of the measure-
In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium – 100 Years ISPRS, Vienna, Austria, July 5–7, 2010, IAPRS, Vol. XXXVIII, Part 7B
Contents
Author Index
Keyword Index
ment probe (see Fig. 4) is a scaled replica of the profile height
h(r).
2 EXPERIMENTAL SETUP
Before a recognition system based on the laser light sectioning
technique is realized, the camera characteristics and furthermore
the depth resolution limits have to be determined. First of all an
approximation of the minimum detectable object size is necessary, therefore the specifications of the provided camera and the
expected viewing geometry must be known:
• Pixel pitch p: 4.65 µm
Figure 3: Scheme for the laser light section technique with the
geometric triangle drawn in produced by the line shaped light
source.
d(r)
= tan(ϕ)
h(r)
h(r) =
d(r)
tan(ϕ)
(1)
(2)
The smallest clearly recognizable profile height depends on the
triangulation angle, ϕ, which is measured between the incidence
direction of the light source and the surveying direction of the
camera and the pixel pitch. For optical reasons (reflection properties of the coil) and resolution considerations the triangulation
angle ϕ should be between 30◦ and 60◦ (Schäfter and Kirchhoff,
2004).
To minimize the influence of the varying lighting conditions a
powerful diode laser with 100 mW at 660 nm is used as light
source and additionally an optical bandpass filter is used to reduce
the ambient lights influence. The optical bandpass (see Fig. 5)
consists of two separate filters, a near infrared blocking filter and
a red filter. The combination of both fiters results in an optical bandpass with a maximum transmission of 94 % at a central
wavelength (CWL) of 685 nm and a full width at half maximum
(FWHM) bandwidth of 100 nm. Compared to a standard interference filter the assembled optical bandpass has a larger FWHM
bandwidth but shows no dependency on the angle of incidence,
which is known to cause spectral side effects for interference filters.
Transmissionfactor T
1
0.8
Red filter
Near infrared filter
Resulting bandpass
0
200
400
600
800
Wavelength in nm
Z
c
x
+1
→
xmin = p
Z
+1
c
(3)
The minimum detectable object size xmin can now be used to
designate the triangulation angle ϕ (in Eqn. 2). In Fig. 6 the relationship between ϕ and the minimum detectable winding displacement hmin is plotted.
Figure 6: Relationship between ϕ and the minimum detectable
winding displacement hmin . The required and realized operating
points for the minimum detectable winding displacement hmin
are marked with red dashed lines. The gray colored area indicates
the interval of technically relevant triangulation angles ϕ. The
green hatched area labels the interval of valid triangulation angles
ϕ for the realization.
The minimum detectable winding displacement hmin was specified to be equal to or less than 1mm. Therefore the triangulation angle ϕ is choosen to be 58◦ , and the resulting minimum
detectable winding displacement hmin is 0.47 mm. Due to the
measurement error which will be calculated in Sec. 5 a value less
than 1mm is advantageous. A major requirement for using the
laser light section technique is the presence of a diffusely reflecting object. This can be proved with the Rayleigh criterion (Horbach, 2008) shown in Eqn. 4.
0.6
0.2
By using Eqn. 3, which describes the central projection theorem (Luhmann, 2003) (with the object dimensions in real world x
and on the CCD chip x′ ), we obtain a minimum detectable object
size xmin of 0.7486 mm.
x′ =
Figure 4: Top view from the laser light section technique scheme.
0.4
• Focal length f : 30 mm (camera constant c ≈ f )
• Distance between camera and coil Z: 5 m
1000
Figure 5: Diagramm of the filter transmittance, the near infrared
blocking filter (green), the red filter (red) and the resulting optical
bandpass (blue).
The second possibility to minimize the influence of the lighting
conditions is to reduce the shutter time. This is also necessary
with regard to the movement of the coil during the measurement.
294
Ra <
λ
16 cos(ϕ)
(4)
The Rayleigh criterion is an inequation including the wavelength
of light λ, the triangulation angle ϕ and the mean roughness index Ra . If the mean roughness index Ra of the illuminated object area is less than λ/16 cos(ϕ) no diffuse reflection appears
In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium – 100 Years ISPRS, Vienna, Austria, July 5–7, 2010, IAPRS, Vol. XXXVIII, Part 7B
Contents
Author Index
according to this law. By evaluating the inequation it follows
that a diffuse reflection appears if Ra ≥77.842 nm. The mean
roughness index Ra of the coil front surface is between 6 µm and
11 µm and therefore it is sufficient to guarantee diffuse reflection.
To proof the assumptions made in this section an experimental
setup has been designed. The coil windings are modelled by several steelplates of the same thickness which are stacked upon each
other and then deferred fixed. First a frontal photograph of the
coil winding model including the laser line was taken to calculate
the displacements with the laser light section technique, shown in
Fig. 7(a). Because of the good illumition conditions only an optical red filter was used. Additionally a photograph from the side
was taken as reference image for the coil winding model, shown
in Fig. 7(b).
500
Height in pixel
Height in pixel
500
1000
1500
2000
1000
1500
Keyword Index
sults in a maximum error of 6 mm for the displacement determination. Futhermore the coil windings can also be affected by a
linear trend due to a non–conform rolling–up process. This linear trend can be used as a quality characteristic too and should
therefore also be measured. To estimate the influence of the laser
alignment and the linear trend of the coil front on the laser line a
model of the setup is introduced in Fig. 9. Here two coordinate
systems are considered, a fixed–place coordinate system given by
x, y, z and a laser coordinate system given by ξ, η, ζ. The laser
coordinate system is rotated by the angles α, β, γ which allow
a free rotation of the laser in space with the corresponding rotation matrices (Cook, 2007) shown in Eqns. 5, 6 and 7. For a
solvable system of equations it is necessary to pose the following
constraint on the coil windings and the laser line. The additional
linear trend resulting from the non–conform rolling–up processes
is represented by a rotation of the coil front around the y–axis
by an angle θ with the corresponding rotation matrix shown in
Eqn. 8. That means that when the laser line is projected onto the
axis of symmetry in the lower half of the coil only a linear trend
is present in the x–direction. This can be achieved by an exact
triggering and a limitation on the ROI.
2000
2500
2500
1000
2000
3000
Width in pixel
(a) Frontal photograph of the
coil winding model showing
the laser line.
1000
2000
3000
Width in pixel
(b) Lateral photograph of the
coil winding model.
Figure 7: Photographs of the coil winding model.
Winding displacement h in mm
To determine the winding displacement, the first step is to extract
the laser line from the acquired image data. Therefore the image
is cropped to the probe edges and then the position of maximum
intensity in every row is determined. For the experimental setup
a commercial photo camera was used, whos images were downsampled to obtain a comparable resolution to typical video cameras. The reference for the coil winding model is obtained from
the lateral image by using a canny edge detection algorithm (Gonzalez and Woods, 2008). In Fig. 8 the results of the previous steps
are shown. The laser light section curve is close to the reference
curve and also the desired resolution limit is almost reached.
10
0
−5
−10
10
20
Profile position in mm
30
Figure 8: The red colored curve is the result for the displacement recognition by using the laser light section technique and
the green colored curve is the reference derived from the frontal
probe image (see Fig. 7(a)). A good agreement can be seen.
3
The following rotation matrices Rγ , Rβ , Rα and Rθ represent
the degrees of freedom in the setup:
• Rotation around ζ, for up or down tilting of the laser.


cos(γ) − sin(γ) 0
cos(γ)
0
Rγ =  sin(γ)
0
0
1
Laser−Light−Section
Reference
5
0
Figure 9: Advanced model of the setup with a fixed–place coordinate system given by x, y, z and a laser coordinate system given
by ξ, η, ζ.
ADVANCED MODELING OF THE SETUP
During the experimental setup the influence of the laser alignment on the result was very low because of the small–sized probe
with a height of only 32 mm (a typical coil profile length can
be up to 700 mm). Considering a profile length of 700 mm a
slight rotation of the laser around its symmetry axis by 0.3◦ , re295
(5)
• Rotation around η, for rotation of the laser around its symmetry axis.


cos(β)
0 sin(β)
0
1
0 
(6)
Rβ = 
− sin(β) 0 cos(β)
• Rotation around ξ, for tangential deviation of the laser due to
the laser light section technique in the following form α =
ϕ − π/2.


1
0
0
cos(α)
sin(α) 
(7)
Rα =  0
0 − sin(α) cos(α)
• Rotation around y, for skewness of the coil front.


cos(θ)
0 sin(θ)
0
1
0 
Rθ = 
− sin(θ) 0 cos(θ)
(8)
In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium – 100 Years ISPRS, Vienna, Austria, July 5–7, 2010, IAPRS, Vol. XXXVIII, Part 7B
Contents
Author Index
With the given rotation matrices the normal vector for the coil
→
front v Cn and the normal vector for the triangular shaped area
→
spanned by the laser v Ln can be calculated in the fixed–place
coordinate system, shown in Eqns 9 and 10.
→
v Ln =
  

0
sin(β)
Rα · Rβ · Rγ · 0 =  sin(α) cos(β) 
cos(α) cos(β)
1
| {z }
(9)
→
eζ
→
v Cn =
  

0
sin(θ)
Rθ · 0 =  0 
1
cos(θ)
| {z }
Keyword Index
Due to Eqn. 13 the gradient of the laser line kyx in the photograph
is dependent on the linear trend of the coil windings represented
by θ and the laser rotation around the symmetry axis represented
by β. Both angles affect the gradient kyx in the same way and
so an estimation using the Eqns 15 and 16 will fail because they
can not be determined seperately by a single measurement. The
solution to this problem is to make a reference measurement for
the skewness of the laser line related to β by using a plane with a
defined angle θ instead of the coil front. A vertical plane as reference (θ = 0 ◦ ) allows to extract β from Eqn. 13, (see Eqn. 17).
(10)
β = − arctan (kyx (θ = 0◦ ) sin(α))
θ = arctan kyx tan(α) + tan(β)
cos(α)
→
eCoil
→
→
→
u = v Ln × v Cn
(11)

xL
 yL  = ν →
u
zL

(12)
To eliminate the influence of the laser alignment and the linear
trend of the coil front the remaining angles β and θ must be calculated. Therefore the orientation of the laser line will be separatlly
examined in the y,x–plane and in the y,z–plane. Due to the setup
the orientation of the laser line kyx in the y,x–plane is present in
the observable image scene in Fig. 9 and can be measured by a
trend estimation of the extracted laser line, the associated relation
of the laser coordinates yL and xL is shown in Eqn. 13.
(18)
Due the fact that the triangulation angle ϕ, and so α (= ϕ−π/2.)
is not constant with respect to β, a slight error occurs. Despite the
fact that ϕ is influenced by β, the approximation approach works
excellent to determine the laser alignment and the linear trend of
the coil front for values of β less than ±15◦ . For example a laser
rotation by β = −7.5◦ results in an error of θ of less than 0.01◦ .
4
LASER LINE EXTRACTION ALGORITHM
Before the coil profil can be determined, the laser line must first
be extracted out of the image. Figure 10 shows a steel coil and a
laser line projected on it. Due to a rotation of the camera of 90◦
(to obtain the profil length at a higher resolution corresponding to
the camera chip with 1392×1040 pixel) the laser line is horizontal in the acquired images. The laser line is clearly visible due to
the use of optical filter mentioned in Sec. 1.
200
Height in pixel
→
Then the direction of the intersection vector u between the coil
front and the laser can be determined by the cartesian product of
→
→
the according normal vectors v Cn and v Ln , shown in Eqn. 11.
Now the intersection of the coil front and the triangular shaped
area spanned by the laser, results in the visible laser line on the
coil with the coordinates (xL , yL , zL ). The intersection vector
→
u is a scaled version (by a factor ν) of the laser line with the
→
same orientation, see Eqn. 12. Further the intersection vector u is
independent of γ so the orientation of the laser line is independent
of an up or down tilting of the laser.
(17)
400
600
800
1000
kyx
cos(α) cos(β) sin(θ) − sin(β) cos(θ)
yL
=
=
xL
sin(α) cos(β) cos(θ)
(13)
The second laser line orientation kyz in the y,z–plane is predetermined by the laser light section technique given by Eqn. 2 and the
associated relation of the laser coordinates yL and zL is shown in
Eqn. 14.
kyz =
cos(α) cos(β) sin(θ) − sin(β) cos(θ)
yL
= tan(ϕ) = −
zL
sin(α) cos(β) sin(θ)
(14)
With Eqns 13 and 14 for the orientation of the laser line it is
possible to determine the remaining angles β and θ, shown in
Eqns 15 and 16.
β = − arctan kyx
cos(α) tan(ϕ)+sin(α)
tan(ϕ)
k
yx
)
θ = − arctan( tan(ϕ)
(15)
(16)
296
200 400 600 800 1000 1200
Width in pixel
Figure 10: Post–processed image (for better contrast of the laser
line) of a steel coil with a laser line.
First of all the algorithm detects if a laser line is present in the image by simply accumulating intensities along rows and searching
for a clearly visible maxima. Also a first elimination of packaged coils and a rough estimation of the laser position is done.
For coils with packaging material a laser line extraction is meaningless and therefore an early elimination is preferable. After
the rough estimation of the laser position, the orignal image is
cropped to a vertical region around the laser. Then a median filter with a rectangular mask (with 15×10 pixel) is applied at the
cropped image to reduce noise and to smooth the laser line, the
result is shown in Fig. 11.
The detection of the maxima positions is an easy way to extract
the whole laser line out of the cropped image. But due to the fact
that the horizontal position of the coils in the images can vary in
dependence on the coil size, the considered image area extends
In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium – 100 Years ISPRS, Vienna, Austria, July 5–7, 2010, IAPRS, Vol. XXXVIII, Part 7B
Author Index
Keyword Index
50
Height in pixel
Height in pixel
Contents
100
150
200
200
400
50
100
150
200
600 800 1000 1200
Width in pixel
200
400
600 800 1000 1200
Width in pixel
Figure 11: Cropped version of the original image after applying
a median filter with differing axes scales.
Figure 12: Laser line located on the coil front extracted with an
adaptive algorithm out of the cropped image.
beyond the coil. So there are also laser lines projected on background geometries present. The background geometries results
in sudden steps of the laser line which can lead to a false detections of profile defects. This problem can be reduced by cropping
the original image to a vertical region around the estimated laser
position. Another problem is the possible presence of edge protection material on the coil that also results in sudden steps of
the laser line. Therefore a criterion to only extract reflections of
the coil front is needed. The following adaptive algorithm is proposed to extract only valid laser lines:
polynomial. Subsequently the linear trend of the coil front can
be calculated using Eqn. 16. The angle θ is important for quality
aspects and also helps to eliminate coils with packaging material,
because the packing material results in nontypical values. For
example a coil with packaging material is shown in Fig. 13(a)
and the extracted laser line with the asymptotic line is shown in
Fig. 13(b). In this case θ is 5.7◦ but it should be less than 1 ◦ for
unpacked coils.
3. Setting of the start point for the adaptive algorithm in the
middle of the laser line to become independent of variing
coil sizes.
4. Setting of the start values for the adaptive algorithm with
DevAv(1) = PosMax (StartPoint).
5. Run through the maxima positions left and right of the start
point and save positions which satisfy the following criterion |PosMax (i + 1) − DevAv(i)| < σ for σ ≤ 1.
6. If the next maximum position satisfies the criterion from
above a new comparison value is calculated for each position with
DevAv(i + 1) = (DevAv(i) + PosMax (i + 1))/2
otherwise the laser line extraction is finished.
Under the assumption (that the alteration of the winding displacements occurs continuously) made in Sec. 1, a restrictive criterion to exclude unvalid steps can be found, see point 5. The local behaviour of the laser line is estimated according to 6 to get
an adaptive limitation. The advantages of this algorithm are in
its simplicity and speed of operation but still delivers good results. In Fig. 12 the result of the algorithm, the extracted laser
line is shown. Alternatively a gray value threshold fiter can be
used to extract the laser line but this has the disadvantage that
unwanted objects are also included. So it is necessary to use
post–processing to adapt the laser line by thinning and to exclude
backgroung geometries and edge protection material.
After the laser line is extracted, the overlayed linear trend represented by kyx can be determined by approximating a first–order
297
Height in pixel
400
600
550
500
800
Asymptotic line
450
Extracted laser line
1000
200 400 600 800 1000 1200
Width in pixel
(a) Post–processed image of a
coil with packaging material
and a laser line.
0
500
1000
Width in pixel
(b) Extracted laser line (blue) and
the corresponding fitted line (red).
Figure 13: Laser line extraction for a coil with packaging material
visible.
By using Eqn. 2 the real coil profile can be calculated out of the
extracted laser line. For the detection of defects the first derivative
of the coil profile is calculated. Then the profile subsegments with
a specified raise related to the deviation and height are marked as
critical. Additonally critical subsegments in a close neighbourhood are joined. The coil profile and the result for the dection of
defects, that are larger than 2 mm, is shown in Fig. 14.
Winding displacement h in mm
2. The maxima positions are normalized to the height of the
search region PosMax (i) = PosMax (i) / Height.
Height in pixel
1. In every row the position of the maximum is detected (under the assumption that only one maximum per column is
present, which is referring to the laser) and then saved in the
variable PosMax (i).
600
200
20
Coil profile
Defects
10
0
−10
−20
200
400
600
Laser position in mm
800
Figure 14: Recognized coil profile (black) and detected defects
(red).
5
RESULTS AND DISCUSSION
To evaluate efficiency of the realized measurement system, the
maximum systematic error ∆h (Schrüfer, 2004) is determined by
In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium – 100 Years ISPRS, Vienna, Austria, July 5–7, 2010, IAPRS, Vol. XXXVIII, Part 7B
Before the total systematic error ∆h can be calculated by Eqn. 19,
the acquisition of the coil profile must be defined. Therefore
Eqn. 2 which describes the calculation of the winding displacement due to the laser light section technique and Eqn. 3 for the
mapping properties due the central projection theorem are used.
So the final function for h is shown in Eqn. 20.
h=
Z
1
+ 1 x′
tan(ϕ)
c
| {z } |
{z
}
laser light
central
section
projection
1
2
∆h = −
∆ϕ · |h| + ∆Z · |h|
sin(2ϕ)
Z +c
1
Z
+ + 1 ∆x′ tan(ϕ) c
400
600
800
1000
10
Coil profile
Defects
Systematic error
0
−10
400
600
Laser position in mm
800
(a) Coil with a laser line and several (b) Final processed croil profile
defects.
(black), defects (red) and systematic error (green).
200
400
600
800
1000
(20)
20
−20
200
200 400 600 800 1000 1200
Width in pixel
Winding displacement h in mm
∂h
∂h
∂h
∆x
∆h = ∆ϕ + ∆Z +
∂ϕ
∂Z
∂x
| {z }
| {z }
| {z }
error due to
error due to
error due to
laser align
camera setup
quantization
(19)
200
Height in pixel
Eqn. 19 as a function of the winding displacement h, the triangulation angle ϕ, the distance Z between camera and coil, and the
quantization of the laser position through the camera.
Keyword Index
Winding displacement h in mm
Author Index
Height in pixel
Contents
200 400 600 800 1000 1200
Width in pixel
20
Coil profile
10
Defects
Systematic error
0
−10
−20
0
200
400
600
Laser position in mm
800
(c) Coil with a laser line and a sin- (d) Final processed croil profile
gle defect in the middle.
(black), defects (red) and systematic error (green).
Figure 15: Results of the measurement system for two different
steel coils.
(21)
elimination of steel coils with packaging material and statistically
evaluate the defect detection rate of the realized system.
ACKNOWLEDGEMENTS
Equation 21 shows the final result for the systematic error ∆h.
Considering the specifications for p, c, Z and ϕ (see Sec. 1) an
uncertainly error of ∆Z =100 mm, ∆ϕ =2 ◦ and ∆x′ = p
results in a maximum systematic error ∆h of 1.462 mm for a
winding displacement h of 5 mm. The main point is that the
terms belonging to the camera setup error and the laser alignment
error which are scaled by the winding displacement h are smaller
than 10−1 and the quantization error is a constant 0.487 mm. The
systematic error ∆h is increasing with nearly h/10 + 0.487 mm.
Additionally the first term in Eqn. 21 shows that the error due the
laser alignment is minimal for triangulation angles ϕ between 30◦
and 60◦ .
In Fig. 15(a) an example for a unpackaged coil with several defects is shown. The final processed coil profile (black), the detected profile defects (red) and the systematic error (green) is
shown in Fig. 15(b). It is observable that the calculated coil profile with an insignificantly small systematic error perfectly corresponds to the laser line in the image and the major defects are
detected. Futhermore a second example for an unpackaged coil
is shown in Fig. 15(c) and processed results in Fig. 15(d). In
the second image additional edge protection material is attached
to the coil (visible on the left side) but is excluded from the coil
profile by the extraction algorithm, as it was demonstrated before.
The authors gratefully acknowledge the partial financial support
for the work presented in this paper by the Austrian Research
Promotion Agency under contract grant 814278 and the Austrian
COMET program supporting the Austrian Center of Competence
in Mechatronics (ACCM). Last but not least the authors thank
the company Industrie Logistic Linz (ILL) for its financial and
technical support.
REFERENCES
Cook, M., 2007. Flight Dynamics Principles. Elsevier.
Gonzalez, R. C. and Woods, R. E., 2008. Digital Image Processing. 3 edn, Pearson Prentice Hall.
Horbach, J., 2008. Verfahren zur optischen 3D-Vermessung
spiegelnder Oberflächen. Universitätsverlag Karlsruhe.
Kraus, K., 1996. Photogrammmetrie - Band2: Verfeinerte Methoden und Anwendungen. Vol. 2, 3 edn, Ferd. Dümmlers Verlag
Bonn.
Luhmann, T., 2003. Nahbereichsphotogrammmetrie — Grundlagen, Methoden und Anwendungen. 2 edn, Wichmann.
6 CONCLUSIONS AND FUTURE WORK
Schäfter and Kirchhoff, 2004. Laser Light Section, a Key Feature
in 3D Laser Measurement Technique. Schäfter and Kirchhoff.
We have shown that a recognition of winding displacements for
steel coils using the laser light section technique can be realized
with a proposed height resolution of less than 1 mm. Furthermore we introduced a mathematical model to eliminate the dependencies from an inaccurate laser alignment and to determine
the linear trend of the coil front as another quality aspect. Finally
we also have shown that the designed laser line extraction algorithm extracts reliably with sufficient speed the laser line belonging to the coil front. The next step is to improve the pre–process
Schrüfer, E., 2004. Elektrische Messtechnik. 8. edn, Hanser Verlag.
298
Wu, P., Yu, F., Li, Z., Yan, Z. and Shun, Y., 1993. Analysis technique for the measurement of a three-dimensional object shape.
Appl. Opt. 32(5), pp. 737–742.